Dynamic model

In robotics, a dynamic robot model typically describes relationships between robot motion and the
forces that cause that motion, so that given one of these quantities, we can determine the other. There
are, therefore, the following problems to be considered.

I. Forward Dynamics: The Forward or direct dynamics problem (FOP) is one where the forces that act
on a robot are given, and we wish to solve for the resulting motion. In its simplest form, the forward
dynamics problem can be expressed symbolically as a vector differential equation of the form
q = h (q, q, 't, manipulator parameters)
where, q is the vector of generalized coordinates Goint variables), q and q are its derivatives with
respect to time, 't is the (input) generalized force vector, i.e., the vector of joint torques and/or joint
forces and the "manipulator parameters" are all those parameters which characterize the particular
geometry and dynamics of a robot manipulator. The importance of forward dynamics in robotics
stems mainly from its use in simulation [10]. Simulation of robot motion is a way of testing control
strategies or manipulator designs prior to the expensive task of working with the actual manipulator.
In general, as we shall see later, equation (1.1.1) is not a simple equation for which an analytic
solution can be provided easily. For a general robot manipulator, equation (1.1.1) is very complex
since it is highly nonlinear with strong coupling between the joint variables. Hence, the Chapter 1:
Introduction 5 solution of (1.1.1) for q requires complex procedures for evaluating h and for
performing numerical integration. Fortunately, a solution for equation (1.1.1) is rarely required in
practical applications. More often, we are interested in the following "inverse" problem. ii) Inverse
Dynamics: The inverse dynamics problem (lDP) is one in which we need to determine the generalized
forces that will produce a speci fied motion trajectory. The inverse dynamics problem can be
described mathematically by an equation of the form 1: = f (q, ci, q, manipulator parameters) (1.1.2)
where, as in (1.1.1), the manipulator parameters describe the particular robot manipulator, 1: is the
vector of the unknown generalized forces and (q, ci, q) is the given manipulator trajectory. Inverse
dynamics is very important in practical robot applications because it enables us to determine the
profile of the generalized forces necessary to achieve a desired robot trajectory. Efficient
computation of the inverse dynamics becomes particularly important when 1: has to be evaluated
online. This can arise in several practical situations, e.g., when the robot payload varies, or when the
desired trajectory has to be modified online (e.g. for collision avoidance). Also, inverse dynamics
plays an important role in many advanced robot control strategies where the inverse dynamics are
used in the feedforward or feedback paths and may need to be computed online [11]. Moreover, to
ensure convergence of the control scheme, the inverse dynamics computations may have to be
performed very frequently. Consequently, the formulation and evaluation of these equations of
motion affect the servo rate of the robot controller and partially determine the feasibility of
implementing many control schemes online. The forward and inverse dynamics problems are two
problems which constitute what is usually known as robot manipulator dynamics [12,13]. However,
since both problems are described by highly nonlinear and 6 Dynamic Analysis of Robot
Manipulators dynamically coupled equations, it can be of great assistance in many robotic
applications if we have available the linearized dynamic equations of a robot manipulator. Thus,
besides forward and inverse dynamics, we may also include the following problem in robot
manipulator dynamics. iii) Linearized Robot Dynamics: As is well known, the linearized dynamics of a
nonlinear system can be described by the following first order vector differential equation (in state-
space form) Ox = A (t)Ox + B (t)ou (1.1.3) where, the matrices A (t) and B (t) are functions of time
and ou and Ox denote small perturbations in the input u and state x, respectively, about some
nominal (given) trajectory. Equation (1.1.3) describes the perturbed motion (for sufficiently small
perturbations) of a dynamical system and is usually derived from the actual nonlinear dynamic
equation (1.1.2) by using a Taylor series expansion about a nominal trajectory [14-16]. The Taylor
series expansion is applicable to nonlinear robot dynamics because, as can be easily shown, the
nonlinearities in robot dynamics are analytic functions of their arguments. Therefore, the derivation
of (1.1.3) from a nonlinear dynamic robot model, at least in principle, does not present any
problems. However, applying the Taylor series expansion to a nonlinear system which has the
complexity of a general robot manipulator dynamics is a challenging problem, especially if one
attempts to derive efficient computational algorithms for determining the coefficient matrices of the
linearized model. Linearized robot dynamics may be used in manipulator control [17]. This is best
illustrated by the following example. In ideal situations, equation (1.1.2) provides the generalized
forces which will drive a manipulator along a desired trajectory. However, in practice, because of
perturbations resulting from modeling errors, unpredicted working conditions or payload variations,
this cannot be achieved without the application of some control strategies, which are designed to
compensate against these perturbations. Currently, Chapter 1: Introduction 7 there are many well
established control strategies in the linear systems area. However, direct application of these linear
control strategies to robotics is not possible, since, as we have already mentioned, the dynamic
models defined by equation (1.1.2) are dynamically coupled and highly nonlinear. Therefore, one
way in which these linear control schemes can be used is by obtaining linearized robot dynamic
models derived from equation (1.1.2). Another application of linearized robot dynamic models is in
carrying out parameter sensitivity analysis of robot manipulator dynamics for the purpose of
efficient manipulator design [14]. 1.2 General Remarks on Robot Manipulator Dyn